You are on page 1of 28

CLASSICAL MECHANICS

UNIVERSIDADE FEDERAL DO RIO GRANDE DO SUL


cucucu
INSTITUTO DE FÍSICA

Profa. Dra. Maria Beatriz de Leone Gay Ducati

April 2021

1 / 28
Symmetry Groups of Mechanical Systems

Canonical transformations form a group, in particular, a Lie Group.

A Lie Group with continuous parameters, θi , has associated with it


a flat vector space whose basis vector, ui , constitute a Lie Algebra
satisfying the Poisson bracket
X
[ui , uj ] = cijk uk . (9.77)
k

The elements Q(θi ), of the associated Lie group are related to the
elements of the Lie Algebra by
i
Q(θi ) = exp( θi ui ). (9.130)
2

2 / 28
Symmetry Groups of Mechanical Systems (cont.)

It might be useful to remember that, by representing the rotation


group in three dimensions, using the hermitian Pauli matrices as
generators of the group, they satisfy the condition

[σi , σj ] = 2iσk ,

for i, j and k a cyclic permutation of x, y and z and


     
0 1 0 −i 1 0
σx = , cucuσy = , cucuσz = .
1 0 i 0 0 −1

For SO(3), then, the structure constant is cijk = 2ijk . Also σi2 = 1 is
the unit 2 × 2 matrix.

3 / 28
Symmetry Groups of Mechanical Systems (cont.)
The Euler angles can be used as parameters that generate the group
elements. For a rotation in the y − z plane, for example, we have

cos 2θ isin 2θ
 
θ θ
Q(θ) = 1cos + iσx sin =  .
2 2 θ θ
isin 2 cos 2

In this formalism, vectors are represented by 2 × 2 matrices of the


form
 
Vz Vx − iVy
V (x, y , z) = ,
Vx + iVy −Vz
and a rotation is performed by a similarity transformation

V (x 0 , y 0 , z 0 ) = Q(θ)V (x, y , z)Q † (θ),

where Q † (θ) is the adjoint of Q(θ).


4 / 28
Symmetry Groups of Mechanical Systems (cont.)

Q are unitary with determinant +1, so they constitute a representa-


tion of the special unitary group in two dimensions, SU(2).

If we also consider the unitary 2 × 2 matrices with determinant −1,


we have the unitary group in two dimensions, U(2).

The last has the same properties as the group of the associated
infinitesimal canonical transformations (I.C.T.).

The Lie Groups of I.C.T’s are known as the symmetry groups of the
system, for such transformations leave the Hamiltonian invariant.

The generators for those groups are the constants of motion.

5 / 28
Symmetry Groups of Mechanical Systems (cont.)

Finding the symmetry groups of the system goes a long way towards
solving it’s classical motion and even closer to a solution of the
quantum-mechanical problem.

A system with spherical symmetry is invariant under rotation about


any axis, so it can be represented by SU(2).

Although, it’s more practical to use the set of usual 3 × 3 rotation


matrix with determinant +1, which represent the special rotation
group in three dimensions, SO(3).

The angular momentum, L is conserved in such a system and its


components are identified as the generators of spatial rotations.

6 / 28
Symmetry Groups of Mechanical Systems (cont.)

The relation [Li , Lj ] = ijk Lk shows that the structure constants are
cijk = ijk and it’s this relationship that stamps the group as being
the rotation group in three dimensions.

The generators of infinitesimal rotations Mi obey the commutation


relations
[Mi , Mj ] = ijk Mk , (4.80)
having the same structure constants as for Li .

The quantities Li and Mi are different physically, but the identity


of said structures constants shows that they have the same group
structure, that of SO(3).

7 / 28
The Bound Kepler Problem
For that problem, together with L, there exists another conserved
quantity A, the Laplace-Runge-Lenz vector, defined by
mkr
A=p×L− . (3.82)
r
Since A qualifies as a system vector, we have

[Ai , Lj ] = ijk Ak . (9.131)

Also, after a tedious manipulation, we have


 
2 2mk
[A1 , A2 ] = − p − L3 . (9.132)
r

The quantity on the right in parenthesis will be recognized as 2mH,


which has the conserved value of 2mE .

8 / 28
The Bound Kepler Problem (cont.)
Introducing
A A
D= √ ≡p , (9.133)
−2mE 2m|E |
then the components of D satisfy

[D1 , D2 ] = L3 .

Thus, the components of L and D together form a Lie Algebra for


the bound Kepler problem, with structure constants to be obtained
from

[Li , Lj ] = ijk Lk , (9.128)


[Di , Lj ] = ijk Dk , (9.134)
and
[Di , Dj ] = ijk Lk . (9.135)

9 / 28
The Bound Kepler Problem (cont.)

An examination of the fundamental matrices for rotation will show


that the symmetry group for the bound Kepler problem is the SO(4),
or the group of the four-dimensional proper rotations.

Such transformations preserve the value of the scalar quadratic form


xµ xµ , where all xµ are real.

An orthogonal transformation in four dimensions has 10 conditions


on the 16 elements of the matrix with determinant ±1, so only 6 are
independent.

Three of the generators are rotations in the various xi − xj planes


and correspond to the Mi from (4.80).

10 / 28
The Bound Kepler Problem (cont.)
Explicitly said generators are given by:
     
0 0 0 0 0 1 0 −1 0
M1 = 0 0 −1 , cuM2 =  0 0 0 , cuM3 = 1 0 0 ,
0 1 0 −1 0 0 0 0 0
except that there are added zeroes in the zeroth row and column.

The remaining three generate infinitesimal rotations in the x0 − xi


planes.

Thus, the generator for an infinitesimal rotation in the x0 − xi planes


would be  
0 −1 0 0
1 0 0 0
N1 =  0 0 0 0 ,
 (9.136)
0 0 0 0
with N2 and N3 given by the same fashion.
11 / 28
The Bound Kepler Problem (cont.)
It should be easy to show that that six matrices satisfy the Lie brackets
[Mi , Mj ] = ijk Mk ,
[Ni , Mj ] = ijk Nk ,
[Ni , Nj ] = ijk Mk ,
with structure constants cijk = ijk .

Since the obtained algebra is the same as in (9.128), (9.134) and


(9.135), the identification of the symmetry group of the bound Kepler
problem as the SO(4) is proven.

For the bound Kepler problem with positive energy (scattering), A is


still a constant of motion , but the appropriate reduced real vector,
instead of D is C defined as
A
C= √ . (9.137)
2mE
12 / 28
The Bound Kepler Problem (cont.)
The Poisson brackets for L and C are now
[Li , Lj ] = ijk Lk , c.
[Ci , Lj ] = ijk Ck , c (9.138)
[Ci , Cj ] = −ijk Lk .
These structure constants are the same as for the restricted Lorentz
group, which must therefore be the symmetry group for the positive
energy Kepler problem - in non relativistic mechanics.

PS.: The fact that the symmetry group involves a space of higher
dimension than ordinary space has nothing to do with special relativity,
but is due to the fact that the symmetry we seek here is one in the
six-dimensional phase space.

The symmetry group consists of the canonical transformations in the


space that leave the Hamiltonian unchanged, so the group can be
interpreted in terms of transformations of spaces of more than 3D.
13 / 28
Two Dimensional Isotropic Harmonic Oscillator
In Cartesian coordinates, the Hamiltonian for this system may be
written as
1 2 1 2
H= (p + m2 ω 2 x 2 ) + (p + m2 ω 2 y 2 ), (9.139)
2m x 2m y
as it does not depend on time explicitly.
The z axis is an axis of symmetry for the system, and hence the
angular momentum along z is a constant of motion:

L = xpy − ypz . (9.140)

Further constants of motion for this problem can be written as


components of a symmetrical two-dimensional tensor A defined as
1
Aij = (pi pj + m2 ω 2 xi xj ). (9.141)
2m

14 / 28
Two Dimensional Isotropic Harmonic Oscillator (cont.)

Of the three distinct elements of the tensor, the diagonal terms may
be identified as the energies associated with the separate
two-dimensional motions along x and y axis, respectively. As there is
no coupling between the two motions, these energies must separately
be constant.

The off-diagonal element of A,


1
A12 = A21 = (px py + m2 ω 2 xy ), (9.142)
2m
is a constant of motion, which can be checked by evaluating the
Poisson bracket with H.

15 / 28
Two Dimensional Isotropic Harmonic Oscillator (cont.)
In relation to the separate x and y motions, A11 and A22 are related
to the amplitudes of the oscillations, whereas A12 is determined by
the phase difference between the two vibrations.

Thus the solutions for the motion can be written as


r
2A11
x= sin(ωt + θ1 ),
mω 2
r
2A22
y= sin(ωt + θ2 ).
mω 2
It follows from (9.142) that
p
A12 = A11 A22 cos(θ2 − θ1 ). (9.143)

The trace of the A tensor is the total energy of the harmonic oscillator.

16 / 28
Two Dimensional Isotropic Harmonic Oscillator (cont.)
Out of the elements of the matrix, we have two other constants of
motion:
A12 + A21 1
S1 = = (px py + m2 ω 2 xy ), (9.144)
2ω 2mω
and
A22 − A11 1
S2 = = (p 2 − px2 + m2 ω 2 (y 2 − x 2 )), (9.145)
2ω 4mω y
To these we may add a third constant of motion from eq. 9.140:
L 1
S3 = = (xpy − ypx ). (9.146)
2 2
The quantities Si plus H form four constants of motion not involving
time explicitly.

Not all of obtained constants of motion can be independent, for in a


system with two degrees of freedom, there can at most be only three
such constants.
17 / 28
Two Dimensional Isotropic Harmonic Oscillator (cont.)
The orbit for the isotropic harmonic oscillator is an ellipse, and three
constants of motion are needed to describe the parameters of the
orbit.
The fourth constant relates to the passage of the particle through a
specific point at a given time and would therefore be explicitly time
dependent.

Hence, there must be a single relation connecting Si and H:

H2
S12 + S22 + S32 = . (9.147)
4ω 2
We can verify that
[Si , Sj ] = ijk Sk . (9.148)
These are the same relations as for the three dimensional angular
momentum vector, thus the group transformation generated by Si
may be identified as SO(3).
18 / 28
Two Dimensional Isotropic Harmonic Oscillator (cont.)
There is an isomorphism between SO(3) and SU(2). It turns out
SU(2) here is more appropriate.

Note that eq. 9.147 suggests there is a three dimensional space, each
point of which corresponds to a particular set of orbital parameters.
For a given system energy, 9.147 says the orbit "points"in this space
lie on a sphere. The constants Si generate three-dimensional rotations
on this sphere; that is, they change the orbit into another orbit having
the same energy.

It may be shown that S1 generates a transformation that changes the


eccentricity of the orbit and that for any given final eccentricity we can
find two transformations leading to it. Is is this double-valued quality
of the transformation that indicates that SU(2) rather than SO(3)
is the correct symmetry group for the two-dimensional harmonic
oscillator.
19 / 28
Two Dimensional Isotropic Harmonic Oscillator (cont.)

For higher dimensions, the structure constants for the Lie Algebras
of SO(n) and SU(n) are no longer identical and a clear separation
between the two can be made.

For the three-dimensional harmonic oscillator, there is again a tensor


constant of motion defined by 9.141, except that now the indices
run from 1 to 3. The distinct components of L now satisfy Poisson
bracket relations with the rather complicated structure constants of
the SU(3) group.

It is possible to show that the n-dimensional harmonic oscillator, the


symmetry group is the SU(n).

20 / 28
Liouville’s Theorem
While the the exact motion of any system is completely determined in
classical mechanics by the initial conditions, it is often impracticable
to calculate an exact solution for complex systems.
In addition, the initial conditions are often only incompletely known.

Statistical mechanics therefore makes no attempt to obtain a com-


plete solution for systems containing many particles. Its aim, instead,
is to make predictions about certain average properties by examining
the motion of a large number of identical systems. The values of the
desired quantities are then computed by forming averages over all
the systems in the ensemble.

Since each system is represented by a single point in phase space, the


ensemble of systems corresponds to a swarm of points is phase space.

Liouville’s theorem states that density of systems in the neighborhood


of some given system in phase space remains constant in time.
21 / 28
Liouville’s Theorem (cont)
The density, D, as defined above, can vary with time by two separate
mechanisms:
There will be an implicit dependence as the coordinates of the system
(qi , pi ) vary with time and also a possible explicit dependence on time.

By means of the equation


du ∂u
= [u, H] + , (9.94)
dt ∂t
the derivative of D due to both types of variation with time can be
expressed as
dD ∂D
= [D, H] + . (9.149)
dt ∂t
The ensemble of system points moving through phase space behaves
much like a fluid in multidimensional space.

22 / 28
Liouville’s Theorem (cont)

In eq. 9.149, the total derivative is a derivative of the density as


we follow the motion of a particular bit of the ensemble "fluid"in time.

On the other hand, the partial derivative is at fixed (q, p); it is as if


we station ourselves in a particular spot in phase space and measure
the time variation of the density as the ensemble of system points
flow by us.

These two derivatives correspond to two viewpoints frequently used


in considering fluid flow.

The total derivative fits in the Lagrangian picture in which individual


particles are followed in time;the coordinates in effect rather identify
a particle than a point in space.

23 / 28
Liouville’s Theorem (cont)
Consider an infinitesimal volume in phase space surrounding a given
system point, with the boundary of the volume formed by some
surface of neighboring system points at the time t = 0.

In the course of time, the system points defining the volume move
about in phase space, and the volume contained by them will take
on different shapes as time progresses (see fig. 1).

Figura: Motion of a volume in two-dimensional phase space.


24 / 28
Liouville’s Theorem (cont)
The number of systems inside the volume remains constant, for a
system initially inside can never get out.

If some system point was to cross the border, it would occupy at


some time the same position in phase space as one of the system
points defining the boundary surface.

Since the subsequent motion of a system is uniquely determined by


its location in phase space at a particular time, the two systems would
travel together from there on. Hence, the system can never leave the
volume. By similar reasoning, no outsider can enter the volume.

The canonical variables (q, p) at time t2 , (fig. 1) are related to the


variables at t1 by a particular canonical transformation. The change
in the infinitesimal volume element about the system point over the
time interval is given by the same canonical transformation.
25 / 28
Liouville’s Theorem (cont)
Now, Poincarè’s integral invariant says that a volume element in
phase space in invariant under canonical transformations. Therefore,
the size of the volume element about the the system point cannot
vary with time.

Thus, both the number of systems in the infinitesimal region, dN,


and the volume, dV , are constants.

As a consequence,
dN
D=
dV
must also be constant in time, that is,
dD
= 0,
dt
which proves Liouville’s theorem.
26 / 28
Liouville’s Theorem (cont)

An alternative statement of the theorem follows from eq. 9.149:


∂D
= −[D, H]. (9.150)
∂t
When the ensemble of systems is in statistical equilibrium, the number
of systems in a given state must be constant in time, which is to say
that the density of system points at a given spot in phase space does
not change with time.

The variation of D in with time at a fixed point corresponds to


the partial derivative with respect to t. Follows, from equilibrium
conditions:

[D, H] = 0.

27 / 28
Liouville’s Theorem (cont)

Given all the above considerations, we can ensure equilibrium by cho-


osing the density D to be a function of those constants of motion of
the system not involving time explicitly, for then the Poisson bracket
with H vanishes.

Thus, for conservative systems D can be any function of the energy,


and the equilibrium conditions are automatically satisfied.

28 / 28

You might also like